See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327720315

Train braking distance calculation using fuzzy logic

Conference Paper · May 2018

CITATIONS READS

2 606

3 authors, including:

Milan Milosavljević
Academy of Technical and Art Applied Studies Serbia, Belgrade
19 PUBLICATIONS   41 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Milan Milosavljević on 18 September 2018.

The user has requested enhancement of the downloaded file.

 TRAIN BRAKING DISTANCE CALCULATION
USING FUZZY LOGIC
Milan Milosavljević*, Dušan Jeremić*, Dušan Vujović *
*
The school of railway applied studies
Zdravka Čelara 14, 11000 Belgrade, Serbia
[email protected], [email protected], [email protected]

Abstract

Train braking distance is one of basic precondition for safety of railway

system. There are many external factors that affect braking distance which can
cause changes in its length. For that purpose, fuzzy logic system based on
experience is proposed.
In this paper, we propose a fuzzy logic system that selects the most likely
control rule from a set of control rules. Input variables are speed, grade,
braking force and braking equipment response time. The aim of this paper is
to determine train braking distance and difference between calculated braking
distance and data from the field. Output variable is provided by a fuzzy logic
system. Fuzzy logic system is simulated using Matlab in fuzzy logic toolbox.
FLS is tested on different train categories and different conditions on the field.

Key words: fuzzy logic, Matlab, train braking distance

INTRODUCTION
Fuzzy logic is a proven methodology for solving the problems with crisp
data, like calculation of braking distance or analyzing the braking performance
which can lead to improvement of the rail lines and promote technical
parameters of trains [1]. This alternative way allows the modeling of a
complex system, using knowledge and experience of the expert and thereby
circumventing the needs for rigorous mathematical calculations.
The aim of the model is reflected in its practical application, i.e. according
to this model it is possible to quickly make the calculation of the required train
braking distance. These obtained values can be used both for newprojected
lines and for checking the train braking distance on the existing lines.
Many papers which uses fuzzy logic for this type of problem, mostly use
input variables for calculation the train braking force and automatic train
control. Main advantage of this fuzzy model in relation to analytical and
simulation models is its simple application in the case when a large number of
input variables do not have a precisely defined value.
 This paper does not refer to the calculation of the train braking distance
and what degree of braking machine engineer should apply, it is primarily
devoted to the design of the railway infrastructure, i.e. those elements whose
layout depends on the length of the train braking distance, such as: signals,
approach contacts, etc.

LITERATURE REVIEW
This section presents a brief description of the relevant literature. There are
a lot of papers that using fuzzy logic for solving problems on railway. The
authors of [2] developed fuzzy logic controller for automatic train control,
using Matlab. Using four input values, they calculate braking system and
torque. Fuzzy logic controller for an automatic train braking system is used in
[3] to develop a braking system that will stop the subway train. Paper [4] gives
the idea of automatic braking system in trains by using artificial intelligence
technique. The intelligence is provided by a fuzzy logic controller, which is
simulated using Matlab-Simulink toolbox. Authors proposed model which can
help in reduction of manpower for train operation on Indian railways.
The authors of [5] presented predictive fuzzy control model for automatic
train operation. In this control model, fuzzy control method and predictive
control algorithm and also computer simulations are all mixed with each other
to compose the predictive fuzzy control.
In paper [6] the general idea of automatic train operation based on
MATLAB. This paper explains Automatic Train Stop and Automatic Warning
System. It is based on the Data Oriented Control Method (DOME) to achieve
train control. Implementation fuzzy logic in solving railway problems is
presented in [7], where authors developed fuzzy model for braking wagon in
marshalling yard.

FUZZY INFERENCE SYSTEM

Standard Fuzzy Inference System (FIS) consist of four elements:
fuzzification, fuzzy rules, conclusion and defuzzification. [8, 9 and 10].
There are two basic FIS types: Mamdani and Takagi-Sugeno. Mamdani’s
approach is characterized by the fact that the final exit from all fuzzy rules is
fuzzy set selected with a minimum strength of rules. Output from Mamdani
type is fuzzy set, which requires aggregation process in the defuzzification
process. Takagi-Sugeno is very similar to Mamdani, there is same approach
in the fuzzification of input variable. The main difference is observed in the
type of output function, which in case of the Sugeno model, only appears as a
linear function or constant [11].
Typical fuzzy logic system is shown on Figure 1 [12]:
 Figure 1. Fuzzy inference system

This rule can be represented as follow:

If x1 is A and x2 is B then y  f x  (1)
where A and B are fuzzy sets in premise, y=f(x) is a faded function in the
sequel. Usually f(x) is a polynomial of constant, but there can be any other
function that adequately describes the system output in the fuzzy domain
defined by the premise if the rule. Output of each rule yi, is hampered by the
strength of the rule wi, which represents the degree of satisfaction of the
premising part of the rule, also known as the degree of fulfilment.
For example, for one “and rule” the strength of the rule is:
w i  AndMethodμAx1 , μBx 2  (2)
where μA(x1) and μB(x2) are membership functions for x1 and x2,
respectively. The final output of the system can be calculated as follow [9]:
N

w y i i
KI  i 1
N
(3)
w
i 1
i

MODEL FORMULATION
In proposed model, there are four input variables, and one output variable.
Input variables are: speed, grade, braking force and braking equipment
response time. This variables were chosen since they have largest impact on
braking distance.
It is known how speed and grade affect the extension of braking distance,
while braking equipment response time and achieved braking force are
variables which depends on many other external parameters. It is not suitable
to use too many input variables due to the large number of possible rules, so
all the others factors which affect on train braking distance are included in
these two variables.
For this kind of problem is developed Mamdani fuzzy model. The model
was tested under normal conditions, i.e. unplanned events such as unbraked
train or braking equipment response time was infinitely large are not
considered.
Membership functions parameters for all variables are given in table 1.

Table 1. Parameters membership functions

Variable Membership function Parameters
(L) Low trampf (0,0,40,70)
x1 - train speed (M) Medium trimf (40,70,100)
(H) High trampf (70,100,120,120)
(LD) Large downgrade trampf (-15,-15,-14,-7)
(SD) Small downgrade trimf (-14,-7,0)
x2 - grade (H) Horizontal trimf (-7,0,7)
(SU) Small upgrade trimf (0,7,14)
(LU) Large upgrade trampf (7,14,15,15)
(S) Small trampf (0,0,0.5,2)
x3 - braking force (M) Medium trimf (0.5,2,3.5)
(B) Big trampf (2,3.5,5,5)
x4 - braking (SH) Short trampf (0,0,1,3)
equipment response (N) Normal trimf (1,3,5)
time (L) Long trampf (3,5,6,6)
300 trampf (0,0,300,500)
500 trimf (300,500,700)
y1 - train braking
700 trimf (500,700,1000)
distance
1000 trimf (700,1000,1100)
1200 trampf (1000,1100,1200,1200)

Train speed
The speed of the train is the parameter of which most depends achievement
certain train braking distance. Braking distance changes exponentially with a
change in speed. On Serbian railway network, maximum speed for all trains
is up to 120 km/h, but average speed is much lower. Division domain of the
input variable „speed” and their membership functions are shown on figure 2.
Membership functions of the triangular and trapezoidal fuzzy numbers „low“,
„medium“ and „high“ are represented in the forms (4)-(6):
 Figure 2. Membership functions for input variable x1 train speed

 1,0  x  40
 70  x
low x  ,40  x  70 (4)
 30
 0, x  70
 0, x  40
 x  40
 30 ,40  x  70

mediumx  1, x  70 (5)
100  x
 30 ,70  x  100
 0, x  100

 0, x  70
 x  70
highx  ,70  x  100 (6)
 30
 1, x  100

Grade
Under normal conditions of exploitation, grade is the second most
important parameter that affect the braking distance. The change in braking
distance with the change of grade also has an exponential characteristics.
Maximum grade on the observed network is up to 18‰, and grade values
greater than 15‰ are present only on very small number of railway sections.
For this reason, the maximum grade value is 15‰.
 low small small large
horizontal
downgrade downgrade  upgrade upgrade

-15-14 -7 0 7 14 15 grade

Figure 3. Membership functions for input variable x2 grade

Braking force
Within the input variable braking force, we include all characteristics of
braking system, adhesion characteristics and external conditions from the field
that can affect on braking distance. When initial braking step is established,
the pressure in the main brake pipe is reduced by 0,5 to 1 bar, and therefore
braking force is small. When fully brake application is established the pressure
in the main brake pipe is reduced by 1 to 3 bar in relations to its normal value
with the release state and therefore braking force is medium. When a process
of rapid or emergency braking occurs, main brake pipe is completely
discharged from 5 to 0 bar and therefore braking force is the largest [13].
The value of the achieved braking force of certain wagons in the train
composition depends on the value of the pressure drop in the main brake pipe,
due to which the triple valve reacts and allows the achievement of a certain
braking force that is proportional to the masses of these wagons
For this reason, the mass of the train has not been taken into consideration
as a separate input variable, it has already been implemented in braking force,
regardless of the train composition.
Membership functions of the triangular and trapezoidal fuzzy numbers
„small“, „medium“ and „big“ are represented in figure 4.
 Figure 4. Membership functions for input variable x3 braking force

Braking equipment response time

Braking equipment response time at higher speeds can have a significant
impact on reaching a certain train braking distance. The assumption is that
only P or R brakes are used. At the time of the response of braking equipment,
the time taken to achieve the maximum value of the braking force under the
observed conditions was taken into account. This time primarily depends on
the length of the train. For the purposes of this paper, the following times of
braking equipment response time were used:
 0-2 s short braking equipment response time,
 2-4 s normal braking equipment response time,
 4-6 s longer braking equipment response time.
Short braking equipment response time is usually achieved with short trains,
while slower response time is present with longer trains and different brake
types in the same train composition.

Figure 5. Membership functions for input variable x4

 Train braking distance
After defining all input variables, we define one output variable, train
braking distance which membership functions of the triangular and
trapezoidal fuzzy numbers „300“, „500“, „700“, „1000“ and „1200“ are
represented in figure 6.
At the observed railway network, at a speed of 100 km/h and on the
horizontal part of the train path, and under normal conditions of exploitation,
the normal value of the braking distance is up to 700 m.

Figure 6. Membership functions for output variable y1 braking distance

FUZZY RULES
After defining all input and output variables, the base of fuzzy rules was
created. One of the most commonly used methods for generating the base of
fuzzy rules is Wang Mendel method, which combines numerical data and
linguistic information [14].
The database contains 135 fuzzy rules, including all combinations of input
variables. All rules are the same weight, with a value of 1. One of the rules is:
If (train speed is high) and (grade is large downgrade) and (braking force
is small) and (braking equipment response time is short) then (train braking
distance is 1200).

RESULTS AND CONCLUSION

Model was tested for real data from the field, for different train categories
and speed. Based on interviews with experts, the values of the input variables
and real train braking distance were recorded. Model results are marked with
y1, and data from field with y1*.
 Table 2. Model results
x1 x2 x3 x4 y1 y1*
100 -10 2 1 1200 1200
100 -10 5 3 979 1000
100 -4 1 2,5 998 1100
80 0 1,5 3 660 700
80 8 1,5 3 446 600
60 -7 3 2 505 500
60 0 5 1 369 400
60 9 5 2,5 268 300
40 -10 1 3,5 427 500
40 0 2 3 201 150

Model results are very close to real data from the field in almost all
situations. We can see that when train speed is very high and other variables
are unsuitable braking distance is close to 1200 m. On the other hand if train
speed is very low, about 40-50 km/h, regardless of the others variables braking
distance is in the range of 200 to 450 m, which fully corresponds to the
situation on the field.
General conclusion is that model gives real data, and in next phase of our
research it can be used for some simulation models. Also model can be
upgraded by using more input parameters, and changing membership
functions of existing variables to be more realistic. Input variable braking
force can be decomposed on new input variables like adhesion characteristics
or detailed external conditions.
In future research and development of this model, authors want to combine
fuzzy logic with simulation modeling of real state, where in various conditions
we can analyze where train stop in stations and in front of signals on open line.

REFERENCES
[1] L. Haidong, “Braking performances of urban rail trains”, Journal of
transportation systems engineering and information technology, vol. 11, issue
6, 2011, pp. 93-97.
[2] M. Madhava at al., “Automatic train control system using fuzzy logic
controller”, Bonfring International Journal of Research in Communication
Engineering, vol. 6, special issue, 2016, pp. 56-61.
[3] M. L. Sharma, S. Atri, “Fuzzy rule based automatic braking system in train
using VHDL”, IJCST, vol. 2, issue 2, 2011, pp. 332-335.
[4] G. Sankar, S. S. Kumar, “Fuzzy logic based automatic braking system in
trains”, International Conference on Power Electronics, 2006, pp. 383-387.
 [5] M. A. Sandidzadeh, B. Shamszadeh, “Improvement of Automatic Train
Operation using enhanced predictive fuzzy control method”, Reliability and
safety in railway, chapter 5, 2012, pp. 121-140.
[6] P.S. Raju, at al., “Automatic Train Operation And Control Using Matlab”,
International Journal of Electrical and Electronic Engineering &
Telecommunications, vol. 2, no. 1, 2013, 150-155.
[7] M. Kapetanović at al., “Razvoj fazi logičkog sistema za upravljanje
kolosečnim kočnicama u železničkoj stanici Beograd ranžirna”, 42nd
International Symposium on Operations Research, 2015, pp. 504-507.
[8] D. Driankov at al., “An Introduction to Fuzzy Control”, Springer-Verlag,
Berlin Heidelberg, Germany, 1993.
[9] G.J. Klir, B, Yuan, “Fuzzy sets and fuzzy logic: theory and applications”,
Prentice-Hall PTR, New York, USA, 1995.
[10] L.X. Wang, “A Course in Fuzzy Systems and Control”, Prentice Hall,
Englewood Cliffs, NJ, USA, 1997.
[11] J. Kiurski-Milošević, “Model procene kvaliteta podzemne vode sa
povećanim sadržajem arsena primenom fazi logike”, PhD thesis, Fakultet
tehničkih nauka, Novi Sad, Serbia, 2015.
[12] B. Dimitrijević, V. Simić, “Neuro-fazi pristup pri proceni broja post
express pošiljaka”, PosTel 2008, Beograd.
[13] V. Vajnhal, “Kočnice i kočenje vozova”, Zavod za novinsko-izdavačku i
propagandnu delatnost JŽ, Belgrade, Serbia, 1991.
[14] D. Teodorović, M. Šelmić., “Računarska inteligencija u saobraćaju”,
Saobraćajni fakultet, Belgrade, Serbia, 2012, ch. 1, sec. 1.1, pp. 9-12.

View publication stats